Information fusion in offspring generation: A case study in DE and EDA

Both differential evolution (DE) and estimation of distribution algorithm (EDA) are popular and effective evolutionary algorithms (EAs) in solving global optimization problems. The two algorithms utilize different kinds of information for generating offspring solutions. In the former, the mutation and crossover operators use the individual information to create trial solutions, while in the later, a probabilistic model is built for sampling new trial solutions, which extracts the population distribution information. It is therefore natural to make use of both kinds of information for generating solutions. In this paper, we propose an algorithm that hybridizes DE and EDA, named as DE/GM, which utilizes both DE crossover/mutation operators and a Gaussian probabilistic model based operator for offspring generation. The basic idea is to generate some of trial solutions by the EDA operator, and to generate the rest by the DE operator. To validate the performance of DE/GM, a test suite of 13 benchmark functions is employed, and the experimental results suggest that DE/GM is promising.